The Agent Skills Directory

[COMMAND_EXECUTION]: The shell script scripts/doctor.sh employs the eval command to run health checks extracted from the WORKSPACE.md file. This pattern allows any string located in the configuration file to be executed directly by the shell without validation.
[REMOTE_CODE_EXECUTION]: The skill is designed to interact with project workspaces that may be cloned from untrusted remote sources. A malicious repository can include a WORKSPACE.md file containing harmful shell commands (such as those for data exfiltration) that will be executed when the user runs the 'doctor' skill.
[PROMPT_INJECTION]: The skill facilitates indirect prompt injection by processing instructions from the WORKSPACE.md file without sanitization.
Ingestion points: The WORKSPACE.md file, specifically the 'Prerequisites' and 'Health Checks' sections.
Boundary markers: None identified; the skill lacks delimiters or instructions to disregard embedded commands.
Capability inventory: Arbitrary shell execution via eval, network access via curl and wget, and local file system access.
Sanitization: No validation, escaping, or filtering is performed on the commands read from the workspace configuration before execution.
[NO_CODE]: The script scripts/doctor.sh references and sources an external utility file lib/workspace-utils.sh which is not included in the provided skill package, making its complete behavior unverifiable.

workspace-doctor